1: \begin{abstract}

2:     The Fr\'{e}chet mean or barycenter generalizes the idea of

3:     averaging in spaces where

4:     pairwise addition is not well-defined. In general metric spaces,

5:     the Fr\'{e}chet sample mean is

6:     not a consistent estimator of the theoretical Fr\'{e}chet mean. 

7:     For graph-valued random variables, for instance, the Fr\'{e}chet

8:     sample mean may fail to converge to a unique value. Hence, it

9:     becomes necessary to consider the convergence of sequences of sets

10:     of graphs. We show that a specific type of almost

11:     sure (a.s.) convergence for the Fr\'{e}chet sample mean previously

12:     introduced by \citet{Ziezold1977} is, in fact, equivalent to

13:     the Kuratowski outer limit of a sequence of Fr\'{e}chet sample means.

14:     Equipped with this outer limit, we provide a new proof of the

15:     strong consistency of the Fr\'{e}chet sample mean for graph-valued

16:     random variables in separable (pseudo-)metric space. Our proof

17:     strategy exploits the fact that the metric of interest is bounded,

18:     since we are considering graphs over a finite number of vertices.

19:     In this setting, we describe two strong laws of large numbers for both

20:     the restricted and unrestricted Fr\'{e}chet sample means of all

21:     orders, thereby generalizing a previous result, due to 

22:     \citet{Sverdrup1981}.

23: \end{abstract}